Tempograms and novelty - Homework week 11; a boxplot on tempo for my 4 playlists


Just for fun (and contrast) I had a look at a song from a favorite Techno DJ of mine, Joris Voorn. The song is Polydub from album Four: https://open.spotify.com/track/7fqAVya5281BFkUokiXAAB?si=44ac45f5d6a94528

As can been seen very clearly: a typical 125 BPM style song. What also can be seen in the above Tempogram is the effect of ‘tempo-octave’ where multiple BPM doubles of 125 BPM can be seen in the graph. It is clear to me that Spotify has a much easier job in detecting a clear BPM pattern in a piece of music which is EDM-derived and has clearly articulated (and metered) percussion elements.



When looking at this boxplot, it strikes me that only the playlist ‘dogs’ has a mean around 78 BPM and the other 3 playlists are higher, even the 2 ‘scientific’ ones. Also striking: the mean (and little SD) in the ‘Humans’ playlist around 120 BPM. I had expected for both science-based lists to have the tempo around 60-80 BPM (for synchronization purposes with humans and dogs rest heartrate) but no such thing; the dogsscience list even has a higher average than the regular dogs list.

When listening to the extreme outliers regarding tempo I can safely assume that the Spotify API isn’t working flawlessly: e.g. the ‘fastest’ song from the dogsscience playlist is “Chorale Prelude No. 5”Ach bleib bei uns Herr Jesu Christ” in B-Flat Major, BWV 649 (Harp Version)” and analysed at 200 BPM by Spotify: https://open.spotify.com/track/2Chgj5q0NgvBlFvzYVJvUa?si=5592371b98464c2d

The same applies to the tempo outlier “Leaves” (https://open.spotify.com/track/60YvUB0fCvLlTDCc99wpgD?si=5f3c0edf44fd4713) for the regular dogs playlist, clocked at 204 BPM. When listening it’s apparently wrong. It’s a piano-only piece of music with clearly some rubato-syle tempo changes.

Initially it was difficult to assess any tempo-related aspects of the song Weightless from tempograms because the lower point of the y-axis was set at 80 BPM whereas Weightless is slower than that. Later on I found out how to set the BPM starting point lower, and set it at 40 BPM.

In addition: I found out (from my own research, i.e. listening to the song on Spotify MANY times and clocking it) that the song Weightless is actually starting a bit slower (namely around 60 BPM) than various sources on the internet (e.g. here) to be starting, which is around 70 BPM. Also, very interestingly, the Spotify API gives a 71 BPM tempo! Listen yourselves, clearly incorrect!

Irrespective of this, the song is intendedly designed to start at a certain BPM and then -after some 5 minutes- slows down towards a lower BPM figure. Reason: assisting in bringing the heartrate of the listener down. Yet, hardly -if at all- to be seen in the tempograms: not the 60 BPM start nor the descend to around 52-54 BPM. The problem relates to the fact that the song does have a (low on loudness but definitely consciously hearable) 4/4 meter-type beat but the chords elements are largely ambient-like synths without clear onsets. There are melodical elements but they play a minor role and don’t necessarily occur in a structured/repetitive manner on anyone of the 4/4 meter beats.

So, I allowed myself a sidestep towards another tool, namely Sonic Visualiser wherein (see blue-pink graph below) the waveform of Weightless is shown together with a green line that shows the result from a transform function (add-on to Sonic Profiler) for tempo and beats. The vertical black line, at 6.02 minutes into the song marks the beginning of the descent of the measured BPM’s. Please note: you see here a so-called ‘tempo-octave’: the green line averages until minute 6.02 around 122 BPM which can be interpreted as double the ‘real’ (at least perceived by me!)